Vision-Based 3D Object Localization Using Probabilistic Models of Appearance Christian Plagemann 1 , Thomas M¨ uller 2 , and Wolfram Burgard 1 1 Department of Computer Science, University of Freiburg, Georges-Koehler-Allee 79, 79110 Freiburg, Germany {plagem, burgard}@informatik.uni-freiburg.de 2 Fraunhofer Institute IITB, Fraunhoferstraße 1, 76131 Karlsruhe, Germany mlt@iitb.fraunhofer.de Abstract. The ability to accurately localize objects in an observed scene is regarded as an important precondition for many practical applications including automatic manufacturing, quality assurance, or human-robot interaction. A popular method to recognize three-dimensional objects in two-dimensional images is to apply so-called view-based approaches. In this paper, we present an approach that uses a probabilistic view- based object recognition technique for 3D localization of rigid objects. Our system generates a set of views for each object to learn an object model which is applied to identify the 6D pose of the object in the scene. In practical experiments carried out with real image data as well as rendered images, we demonstrate that our approach is robust against changing lighting conditions and high amounts of clutter. 1 Introduction In this paper, we consider the problem of estimating the three-dimensional posi- tion and the orientation of rigid objects contained in images. This problem has been studied intensively in the computer vision community and its solution is regarded as a major precondition for many practical applications, like automatic manufacturing, quality assurance, or human-robot interaction. In this work, we are especially interested in view-based approaches, where objects are represented by 2-dimensional views. Such approaches allow to incorporate visual features di- rectly and do not assume prior knowledge about the spatial structure of the objects. The limited localization accuracy caused by the view-based representa- tion can be compensated for by a scene-based object tracking process, as will be demonstrated in Section 5.2. Recently, Pope and Lowe [1] proposed the probabilistic alignment algorithm to identify two-dimensional views of objects in images. The goal of the work presented here is to investigate how this purely image-based approach can be utilized to achieve a robust estimate of the position and orientation of the ob- ject in the scene. The input to our system are either real images of an object or alternatively a volumetric model that is used to render the necessary views. We describe how the four parameters obtained from the 2D object recognition can W. Kropatsch, R. Sablatnig, and A. Hanbury (Eds.): DAGM 2005, LNCS 3663, pp. 184–191, 2005. c  Springer-Verlag Berlin Heidelberg 2005